Annotation-free method segments retinal vessels with strong accuracy
FPG-Seg: Annotation-free Retinal Vessel Topology Segmentation
Computer Vision and Pattern Recognition
Summary
Segmenting the tiny blood vessels in retinal images typically needs expert-made labels, which take a lot of time to create. The authors present FPG-Seg, a new method that segments these vessels without any manual labels by combining traditional image filtering, certain deep learning techniques, and graph methods that focus on vessel structure. Their method performs nearly as well as fully supervised models that require labeled data when tested on several common datasets. This approach could speed up analysis in eye health without the usual slow annotation step.
What this means in practice
- •For medical imaging developers: Develop software for retinal vessel analysis that does not require costly manual annotation for training segmentation models.
- •For eye care clinics: Integrate automated vessel segmentation tools that speed up retinal imaging analysis for diagnostics with reduced manual labeling requirements.
Authors
James Willoughby, Irina Voiculescu
Abstract
State-of-the art approaches to retinal vessel segmentation typically require extensive annotations, a highly time-consuming process, reliant on clinical expertise. We present FPG-Seg, an annotation-free pipeline for retinal vessel segmentation combining classical vesselness filtering, Bayesian deep learning and topology-aware graph reconstruction to segment retinal vessel topology without manual labels. Tested across four widely used benchmark datasets for retinal vessel segmentation, FPG-Seg attains a clDice of $0.74{-}0.78$, representing $86{-}98\%$ of the performance of an equivalent fully-supervised baseline, despite using no manual annotations for training or hyperparameter tuning. As part of this work, we also provide, to our knowledge, the first performance benchmark of an annotation-free vessel segmentation method on FIVES.